| blob | 6e048f37ba6542252a8433e1a7e21baa1e51bf61 |
1 2011-02-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
3 * I think interval guaranteed with the $last_aggregate_call
4 variable (both in fsck and server) need to be fixed with a $recc
5 variable to something like 60 seconds or even less.
7 * minor bug: fsck just added the lockfile to the index which should be
8 considered bookkeeping.
10 * speed: a large directory remove that is done by the kernel in 1
11 seconds triggers minutes of bookkeeping (07:56:42 - 08:01:10 for ~3000
12 files, all due to Schlemiehl again. Want to do some collecting of
13 pending ops before actually writing to the RECENT files. Lock!
15 Repeating the timing with 2600 files and it took 21:28:08 - 21:35:27.
17 Now having rewritten the loop to use batch_update(): 4383 file removed
18 21:43:37 - 21:44:12. From 7:30 to 0:35 while doing 60% more work that's
19 a joy.
21 * lockdirectory expiration? server died and blocked fsck for so long.
23 * At the moment rrr-init is not needed anymore, rrr-server can be
24 started and followed by rrr-fsck. Still stupid but not the main
25 showstopper.
27 2011-02-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
29 * IN_Q_OVERFLOW needs to trigger an fsck, likewise the entry into the
30 server.
32 2011-01-30 Andreas J. Koenig <andk@cpan.org>
34 * rrr-init and rrr-server need to be used such:
36 init
37 server
38 init again
40 Stupid but no showstopper ATM.
42 2010-10-26 Andreas J. Koenig <andk@cpan.org>
44 * forking slave is done, needs more testing.
46 2010-07-06 Andreas J. Koenig <andk@cpan.org>
48 * The two next most important goals are to get a useable rrr-server and
49 a forking slave implementation that uses virtually no memory after the
50 first pass through. Does it need the status file?
52 * rrr-init now being fast we can now remove the alpha alerts. Done.
54 2010-06-16 Andreas J. Koenig <andk@cpan.org>
56 * rrr-init seems to be sssslllloooowwww. At 3:49 I started it for the
57 video tree. After twelve minutes the file has reached a size of 500k. It
58 is locking and rewriting the YAML after every file. Wasn't there an
59 unlocked variant? For "init" one would not expect locked correctness and
60 a rewrite after every item. Given that this tends to become slower with
61 every file it is really unacceptable.
63 Reminds me of rrr-dirtyupdate (the former name of rrr-update). It seems
64 the code was not dirty enough and that was the reason why it was so
65 slow. And rrr-update does not have a --dirty switch, so it looks like we
66 have no way to force dirty and fast operation yet but we need it for
67 init.
69 Irritating the bootstrapping race: the order {(1) fill initial
70 array, (2) start server} leaves the need for an fsck to cover the time
71 between 1 and 2. Given that regular fsck is needed in any case the price
72 is not high but still doesn't give a warm feeling. Maybe starting the
73 server will by itself do an fsck?
75 2010-06-15 Andreas J. Koenig <andk@cpan.org>
77 * remove the wild alpha alerts and say, it is alpha but parts of it are
78 already working very well or so.
80 2009-12-10 Andreas J. Koenig <andk@cpan.org>
82 * Prominent request for statusfile to safe memory consumption. I'd like
83 to work backwards: get a verification program that verifies how well in
84 sync two nodes are without using frmr itself. Pretend that the forking
85 agent already works and can be interrupted with ^C. Then let it die at
86 random and restart at random and see the memory consumption and the
87 failure rate over time. Compare it with the curve we have now.
89 Of course we also invented the status file as a sharp debugging tool.
90 I'd like to see it in action. See also: F:R:M:Recent::thaw().
92 2009-06-22 Andreas J. Koenig <andk@cpan.org>
94 * Todo: rrr-init to generate some index files so that rrr-server can be
95 started.
97 2009-05-02 Andreas J. Koenig <andk@cpan.org>
99 * DONE: rename rrr-dirtyupdate to rrr-update and add an option --dirty
100 so we can use it for normal update with a single file. And add a cronjob
101 that runs rrr-fsck so we get a reminder when something is not working
102 as it should.
104 2009-04-27 Andreas J. Koenig <andk@cpan.org>
106 * Todo: I still see more link_stat errors than expected. I come back to
107 the former suspicion that we do not do necessary bookkeeping on deleted
108 files so that when we reach the "new" event on a meanwhile deleted file
109 we try to mirror it although we could know better. It's just noise, no
110 harm but noise is irritating.
112 * Todo: more experimenting with runstatusfile to get it official
114 * Todo: make sure $ENV{LANG}="C" when we call rsync because we will
115 parse the error.
117 rsync: link_stat "/authors/id/T/TH/THINC/DateTime-Format-Flexible-0.02.tar.gz" (in PAUSE) failed: No such file or directory
119 rsync: link_stat "/authors/id/A/AG/AGENT/SSH-Batch-0.001.meta" (in PAUSE) failed: No such file or directory (2)
121 And that ignore_link_stat_error is propagated in sparse_clone.
123 DONE
125 For the record: the bug was that ignore_link_stat_error was lost during
126 sparse_clone. And then the second bug was that it did not default to
127 true. Both are now fixed.
129 2009-04-25 Andreas J. Koenig <andk@cpan.org>
131 * Todo: fill rrr program with life
133 2009-04-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
135 * Minor bug: I see complaints about files not existing like
136 /authors/id/G/GW/GWADEJ/SVG-Sparkline-0.2.0.meta or
137 /authors/id/R/RC/RCAPUTO/POE-1.004.readme
139 In the RECENT files these are delete events but of course also new
140 events in older recent files.
142 I think we do not properly remember deletes when we skip-deletes and so
143 do not filter them out when we later see the "new" event.
145 Yes, but that what ignore_link_stat_error is for, we know about this
146 race condition. We must make ignore_link_stat_error default to true.
148 DONE
150 * bug in Done.pm failing to merge two fields into one when a third is
151 present. FIXED and accompanied with new test.
153 2009-04-24 Andreas J. Koenig <andk@cpan.org>
155 * Todo: keep the index clean and run some kind of nondestructive fsck 4
156 times a day.
158 * who is our backbone?
160 http://cpan.cpantesters.org/authors/02STAMP # barbie
161 http://cpan.solfo.com/authors/02STAMP # abh
162 (http|ftp)://theoryx5.uwinnipeg.ca/pub/CPAN/ # rkobes
164 2009-04-17 Andreas J. Koenig <andk@cpan.org>
166 * the changelog helper I included in the Makefile release memo does not
167 show tags and is much less useful than I thought. gitk is probably much
168 more convenient.
170 2009-04-16 Andreas J. Koenig <andk@cpan.org>
172 * ABH writes:
174 #### Sync 1239828608 (1/1/Z) temp .../authors/.FRMRecent-RECENT-Z.yaml-
175 #### Ydr_.yaml ... DONE
176 #### Cannot stat '/mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-
177 #### Ydr_.yaml': No such file or directory at /usr/lib/perl5/site_perl/
178 #### 5.8.8/File/Rsync/Mirror/Recentfile.pm line 1558.
179 #### unlink0: /mirrors/CPAN/authors/.FRMRecent-RECENT-Z.yaml-Ydr_.yaml is
180 #### gone already at cpan-pause.pl line 0
182 Running without skip-deletes now but cannot reproduce.
184 2009-04-15 Andreas J. Koenig <andk@cpan.org>
186 * consider state directory such we can restart after a ^C or come again
187 from a cronjob.
189 Reuse _runstatusfile: write atomically, see how we can use the dump to
190 start running again. Do not leak and remove globs before writing. Do not
191 only dump one rf, dump the whole complex.
193 * consider the effect of ^C. Is it safe?
195 * want to use Perl::Version to up the versions on every release. Only if
196 things have changed, of course. Does this actually work?
198 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun
199 /usr/local/perl-5.10-g/bin/perl-reversion -dryrun -bump lib/File/Rsync/Mirror/Recentfile.pm
201 Must do the bookkeeping myself when this should happen, maybe based on git?
203 I think I prefer the same version number for all pm files. But one also
204 needs a routine for setting them. Ahh:
206 /usr/local/perl-5.10-g/bin/perl-reversion -current 0.0.4 -set 0.0.5 lib/**/*.pm
208 NEEDSMOREWORK
210 * why indexer wrong?
212 cpan[2]> m /Mirror::Recent/
213 Module = File::Rsync::Mirror::Recent (ANDK/File-Rsync-Mirror-Recent-0.0.4.tar.bz2)
214 Module = File::Rsync::Mirror::Recentfile (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
215 Module = File::Rsync::Mirror::Recentfile::Done (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
216 Module = File::Rsync::Mirror::Recentfile::FakeBigFloat (ANDK/File-Rsync-Mirror-Recent-0.0.2.tar.bz2)
217 4 items found
219 The META.yml was generated by ExtUtils::MakeMaker version 6.5101
221 The 0.0.2 release has a META.yml by ExtUtils::MakeMaker version 6.42 and without "provides".
223 I just tried MM 6.5102 and it again produces only a provides for Recent,
224 not for the other modules.
226 My own fault. FIXED in the Makefile.PL.
228 2009-04-13 Andreas J. Koenig <andk@cpan.org>
230 * Study http://en.wikipedia.org/wiki/Magnet_URI_scheme
232 * is it true (as stated at
233 https://fedorahosted.org/InstantMirror/wiki/ExistingRepositoryReplicationMethods)
234 that rsync can lead to errors when upstream changes before client sync
235 has completed? Can we see the error when we run this on upstream:
237 % perl -e '
238 use Time::HiRes qw(time);
239 while (){
240 open my $fh, ">", "changingfile.txt.new" or die;
241 print $fh (time."\n") x 1000000;
242 close $fh;
243 rename "changingfile.txt.new", "changingfile.txt" or die;
244 }
245 '
247 And on the receiving end:
249 % while true; do
250 rsync k75:`pwd`/changingfile.txt .; cat changingfile.txt| uniq|wc -l
251 sleep 1;
252 done
254 I cannot get it to fail. Apparently it is sufficient when upstream
255 always writes atomically (which of course is mandatory)?
257 SENT to mailing list Apr 28.
259 * from ABH:
261 > https://fedorahosted.org/InstantMirror/
262 > https://www.redhat.com/mailman/listinfo/instantmirror-list
263 > irc.freenode.net channel #instantmirror
265 A cool name for a project. Inspires me to write a few sentences about
266 bittorrent's role in the grand picture.
268 http://spreadsheets.google.com/pub?key=pGlWX10blP4u2kM05SDtiMg is a
269 spreadsheet collected by Atul Aggarwal about bittorrent implementations.
271 2009-04-12 Andreas J. Koenig <andk@cpan.org>
273 * Interesting last minute bug during real download testing: the output
274 isn't as pretty anymore, it seems that more work is being done than
275 needed.
277 Introducing a highlevel _alluptodate method helped a lot in debugging
278 this.
280 FIXED now. Reason was that dirtymark on PAUSE currently is broken; it's
281 different on 1h file than on the other files. This lead to frequent
282 calls to _fullseed. Limiting the dirtymark check to $i==0 resolved the
283 issue.
285 2009-04-12 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
287 * Todo: try if we can be faster with a native float package. I'm really
288 glad to have a machine exposing perl floating point bugs, it forced me
289 to retract this item from the changes file: "several speedups in the
290 fakebigfloat code" before I get cpan testers fails.
292 2009-04-11 Andreas J. Koenig <andk@cpan.org>
294 * Bug: when the dirtymark changes upstream, then the downstream server
295 notices it immediately and mirrors the "1h" file but then leaves the
296 rmirror loop and continues with "6h" on the next round through. This
297 behaviour goes on until it reaches the Z file. Because we eat the Z file
298 piecemeal we leave the loop after a while but after that everybody seems
299 to have forgotten that there is some work left to be done. Bad, bad,
300 bad.
302 ETOOCONFUSING better name?
304 Recentfile::get_remote_recentfile_as_tempfile OK
305 Recentfile::resolve_recentfilename split_rfilename DONE
307 Recent::_principal_recentfile_object _principal_recentfile_fromremote
308 Recent::principal_recentfile OK
309 Recent::_resolve_rfilename _principal_recentfile_fromremote_resosymlink
311 some void, some not, some not void but called in void context.
313 2009-04-10 Andreas J. Koenig <andk@cpan.org>
315 * Bug? When the dirtymark changes then the second tier hosts quickly
316 reset their DONE state and restart mirroring. But then they mirror
317 potentially outdated index files with inconsistent dirtymark because the
318 first tier box may be a bit behind on the large recentfiles.
320 This means we need a brake when iterating over the recentfiles that
321 refuses to use a recentfile with the wrong dirtymark.
323 2009-04-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
325 * install the bin/* files and talk about rrr-overview somewhere.
327 * Ask B. Hansen asks for tempdir accessor such that tempfiles get
328 created outside the target tree.
330 * Barbie asks for better logging capabilities
332 2009-03-29 Andreas J. Koenig <andk@cpan.org>
334 * known bug: cannot configure the Recentfile objects to keep delete
335 events. How to call the accessor for that? keep_delete_objects_forever?
337 Sounds acceptable. FIXED (but untested).
339 * known bug: keeps temporary index files lying around.
341 2009-03-24 Andreas J. Koenig <andk@cpan.org>
343 * Some equivalent for
345 for ( @{$rrr->recentfiles} ) { $_->verbose(0) }
347 ? DONE
349 2009-03-22 Andreas J. Koenig <andk@cpan.org>
351 * k81 is again client of k75 with different parameters.
353 * broken now: (1) seeding and unseeding: rmirror is seeding and talking
354 about it all the time and nobody reacts accordingly; (2) lots of temp
355 files get created and not removed; culprit the new call to
356 get_remote_recentfile_as_tempfile within Recent.pm; the manpage says the
357 caller has to remove the tempfile after use.
359 Need the drawing board.
361 PARTLY FIXED: retracting the idea to call
362 get_remote_recentfile_as_tempfile from Recent.pm and moving around seed
363 and unseed such that uptodate now means "we have mirrored this
364 recentfile's bunch of files" and seeded means "somebody believes the
365 index file for this rf needs be refreshed". This seems to work now.
367 Still collecting temp files that nobody cares to depollute.
369 2009-03-21 Andreas J. Koenig <andk@cpan.org>
371 * Bug?: should it be harder than it is atm to set the timestamp to the
372 future? YES, FIXED
374 * bug with native integers:
376 -
377 epoch: 997872011
378 path: id/D/DE/DELTA/Crypt-Rijndael_PP-0.03.readme
379 type: new
380 -
381 epoch: 1195248431
382 path: id/L/LG/LGODDARD/Tk-Wizard-2.124.readme
383 type: new
385 Native integer broke when native math was turned off. FIXED
387 * Bug: something between id/P/PH/PHISH/CGI-XMLApplication-1.1.2.readme
388 and id/C/CH/CHOGAN/HTML-WWWTheme-1.06.readme. Mirroring the Z file loops
389 in this area.
391 Yes, records out of order:
393 447003 -
394 447004 epoch: 995885533
395 447005 path: id/P/PH/PHISH/CGI-XMLApplication_0.9.3.readme
396 447006 type: new
397 447007 -
398 447008 epoch: 995890358
399 447009 path: id/H/HD/HDIAS/Mail-Cclient-1.3.readme
400 447010 type: new
401 447011 -
402 447012 epoch: 995892221
403 447013 path: id/H/HD/HDIAS/Mail-Cclient-1.3.tar.gz
404 447014 type: new
406 FIXED with sanity check and later with the integer fix.
408 * Bug: want the index files in a .recent directory
410 * Bug: lots of dot files are not deleted in time
412 * possible test case: can a delete change the timestamp? This would
413 probably break the order of events.
415 2009-03-20 Andreas J. Koenig <andk@cpan.org>
417 * 1233701831.34486 what's so special about this number/string? It
418 belongs to
419 id/G/GR/GRODITI/MooseX-Emulate-Class-Accessor-Fast-0.00800.tar.gz and
420 atm lives in Y,Q, and Z.
422 It is the first entry after id/--skip-locking which has timestamp
423 1234164228.11325 which represents a file that doesn't exist anymore.
425 * another bug:
427 Sync 1237531537 (31547/33111/Z) id/J/JH/JHI/String-Approx-2.7.tar.gz ...
428 _bigfloatcmp called with l[1237505213.21133]r[UNDEF]: but both must be defined at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
429 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
430 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
431 File::Rsync::Mirror::Recentfile::Done::covered('File::Rsync::Mirror::Recentfile::Done=HASH(0x8857fb4)', 1237505213.21133, 0.123456789) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 2041
432 File::Rsync::Mirror::Recentfile::uptodate('File::Rsync::Mirror::Recentfile=HASH(0x8533a2c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 536
433 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x82ef3d0)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
434 at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
435 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 131
436 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatlt(1237505213.21133, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 110
437 File::Rsync::Mirror::Recentfile::Done::covered('File::Rsync::Mirror::Recentfile::Done=HASH(0x8857fb4)', 1237505213.21133, 0.123456789) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 2041
438 File::Rsync::Mirror::Recentfile::uptodate('File::Rsync::Mirror::Recentfile=HASH(0x8533a2c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 536
439 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x82ef3d0)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
441 FIXED, it was the "--skip-locking" file where manual intervention was participating
443 * bug on the mirroring slave: when the dirtymark gets increased we
444 probably do not reset the done intervals. The mirrorer stays within
445 tight bounds where it tries to sync with upstream and never seems to
446 finish. In the debugging state file I see lots of identical intervals
447 that do not get collapsed. When I restart the mirrorer it dies with:
449 Sync 1237507989 (227/33111/Z) id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz ...
450 _bigfloatcmp called with l[1237400817.94363]r[UNDEF]: but both must be defined at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 76
451 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatcmp(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/FakeBigFloat.pm line 101
452 File::Rsync::Mirror::Recentfile::FakeBigFloat::_bigfloatge(1237400817.94363, undef) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 226
453 File::Rsync::Mirror::Recentfile::Done::_register_one('File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'HASH(0xb693f2dc)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile/Done.pm line 200
454 File::Rsync::Mirror::Recentfile::Done::register('File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'ARRAY(0x8c54bfc)', 'ARRAY(0xb67e618c)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 1044
455 File::Rsync::Mirror::Recentfile::_mirror_item('File::Rsync::Mirror::Recentfile=HASH(0x84abcb0)', 227, 'ARRAY(0x8c54bfc)', 33110, 'File::Rsync::Mirror::Recentfile::Done=HASH(0x84c6af8)', 'HASH(0x84abd8c)', 'ARRAY(0x839cb64)', 'HASH(0x839c95c)', 'HASH(0xb6a30f0c)', ...) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recentfile.pm line 992
456 File::Rsync::Mirror::Recentfile::mirror('File::Rsync::Mirror::Recentfile=HASH(0x84abcb0)', 'piecemeal', 1, 'skip-deletes', 1) called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 564
457 File::Rsync::Mirror::Recent::_rmirror_mirror('File::Rsync::Mirror::Recent=HASH(0x84ab6d4)', 7, 'HASH(0x8499488)') called at /home/k/sources/rersyncrecent/lib/File/Rsync/Mirror/Recent.pm line 532
458 File::Rsync::Mirror::Recent::rmirror('File::Rsync::Mirror::Recent=HASH(0x84ab6d4)', 'skip-deletes', 1) called at /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl line 27
460 and debugging stands at
462 At: before
463 Brfinterval: Z
464 Ci: 227
465 Dre+1:
466 epoch: 1237400802.5789
467 path: id/J/JO/JOHND/CHECKSUMS
468 type: new
469 Dre-0:
470 epoch: 1237400807.97514
471 path: id/M/MI/MIYAGAWA/CHECKSUMS
472 type: new
473 Dre-1:
474 epoch: 1237400817.94363
475 path: id/X/XI/XINMING/Catalyst-Plugin-Compress.tar.gz
476 type: new
477 Eintervals:
478 -
479 - 900644040
480 - 900644040
481 - []
483 and it is reproducable.
485 Why does the mirrorer not fetch a newer Z file? It is 18 hours old while
486 pause has a fresh one.
488 FIXED, it was the third anded term in each of the ifs in the IV block in
489 _register_one: with that we make sure that we do not stamp on valuable
490 interval data.
492 2009-03-17 Andreas J. Koenig <andk@cpan.org>
494 * done: verified the existence of the floating point bug in bleadperl
495 and verified that switching from YAML::Syck to YAML::XS does not resolve
496 it.
498 BTW, the switch was doable with
500 perl -i~ -pe 's/Syck/XS/g' lib/**/*.pm t/*.t
502 and should be considered as a separate TODO
504 * todo: integrate a dirty update with two aggregate calls before
505 unlocking for frictionless dirtying DONE
507 * todo: start the second rsync daemon on pause DONE
509 * todo: move index files to .recent: this cannot simply be done by
510 setting filenameroot to .recent/RECENT. Other parts of the modules rely
511 on the fact that dirname(recentfile) is the root of the mirrored tree.
513 2009-03-16 Andreas J. Koenig <andk@cpan.org>
515 * What was the resolution of the mirror.pl delete hook bug? Do we call
516 the delete hook when pause removes a file from MUIR?
518 * Today on pause: Updating 2a13fba..29f284d and installing it for
519 /usr/local/perl-5.10.0{,-RC2}
521 TURUGINA/Set-Intersection-0.01.tar.gz was the last upload before this
522 action and G/GW/GWILLIAMS/RDF-Query-2.100_01.tar.gz the first after it
524 2009-03-15 Andreas J. Koenig <andk@cpan.org>
526 * currently recent_events has the side effect of setting dirtymark
527 because it forces a read on the file. That should be transparent, so
528 that the dirtymark call always forces a cache-able(?) read.
530 * The bug below is -- after a lot of trying -- not reproducible on a
531 small script, only in the large test script. The closest to the output
532 below was:
534 #!perl
535 use strict;
536 use Devel::Peek;
537 my $x = "01237123229.8814";
538 my($l,$r);
539 for ($l,$r) {
540 $_ = "x"x34;
541 }
542 ($l,$r) = ($1,$2) if $x =~ /(.)(.+)/;
543 $r = int $r;
544 $l = "1237123231.22458";
545 $r = "1237123231.22458";
546 1 if $l/1.1;
547 Devel::Peek::Dump $l;
548 Devel::Peek::Dump $r;
549 Devel::Peek::Dump $x = $l <=> $r;
550 die "BROKE" if $x;
551 __END__
553 The checked in state at c404a85 fails the test with my
554 /usr/local/perl-5.10-uld/bin/perl on 64bit but curiously not with
555 /usr/local/perl-5.10-g/bin/perl. So it seems the behaviour is not even
556 in the test script always consistent.
558 * Todo: write a test that inserts a second dirty file with an already
559 existing timestamp. DONE
561 * Bug in perl 5.10 on my 64bit box:
563 DB<98> Devel::Peek::Dump $l
564 SV = PVMG(0x19e0450) at 0x142a550
565 REFCNT = 2
566 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
567 IV = 0
568 NV = 1237123231.22458
569 PV = 0x194ce70 "1237123231.22458"\0
570 CUR = 16
571 LEN = 40
573 DB<99> Devel::Peek::Dump $r
574 SV = PVMG(0x19e0240) at 0x142a3e8
575 REFCNT = 2
576 FLAGS = (PADMY,POK,pPOK)
577 IV = 1237123229
578 NV = 1237123229.8814
579 PV = 0x19ff900 "1237123231.22458"\0
580 CUR = 16
581 LEN = 40
583 DB<100> Devel::Peek::Dump $l <=> $r
584 SV = IV(0x19ea6e8) at 0x19ea6f0
585 REFCNT = 1
586 FLAGS = (PADTMP,IOK,pIOK)
587 IV = -1
589 DB<101> Devel::Peek::Dump $l
590 SV = PVMG(0x19e0450) at 0x142a550
591 REFCNT = 2
592 FLAGS = (PADMY,NOK,POK,pNOK,pPOK)
593 IV = 0
594 NV = 1237123231.22458
595 PV = 0x194ce70 "1237123231.22458"\0
596 CUR = 16
597 LEN = 40
599 DB<102> Devel::Peek::Dump $r
600 SV = PVMG(0x19e0240) at 0x142a3e8
601 REFCNT = 2
602 FLAGS = (PADMY,NOK,POK,pIOK,pNOK,pPOK)
603 IV = 1237123231
604 NV = 1237123231.22458
605 PV = 0x19ff900 "1237123231.22458"\0
606 CUR = 16
607 LEN = 40
609 Retry with uselongdouble gives same effect. Not reproducable on 32bit box (k75).
611 * Todo: reset "done" or "covered" and "minmax" after a dirty operation? DONE
613 2009-03-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
615 * $obj->merge ($other) needs to learn about equal epoch which may happen
616 since dirty_epoch intruded.
618 * Wontfix anytime soon: I think we currently do not support mkdir. Only
619 files!
621 2009-01-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
623 * Todo: continue working on update(...,$dirty_epoch). It must be
624 followed by a fast_aggregate! DONE
626 2008-12-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
628 * maybe we need a closest_entry or fitting_interval or something like
629 that. We want to merge an event into the middle of some recentfile.
630 First we do not know which file, then we do not know where to lock,
631 where to enter the new item, when and where to correct the dirtymark.
633 So my thought is we should first find which file.
635 Another part of my brain answers: what would happen if we would enter
636 the new file into the smallest file just like an ordinary new event,
637 just as an old event?
639 (1) we would write a duplicate timestamp? No, this would be easy to
640 avoid
642 (2) we would make the file large quickly? Yes, but so what? We are
643 changing the dirtymark, so are willing to disturb the downstream hosts.
645 2008-11-22 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
647 * 10705 root 17 0 725m 710m 1712 S 0.0 46.8 834:56.05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
649 leak!
651 https://rt.cpan.org/Ticket/Display.html?id=41199
653 * bzcat uploads.csv.bz2 | perl -F, -nale '$Seen{$F[-1]}++ and print'
655 Strangest output being HAKANARDO who managed to upload
657 Here is a better oneliner that includes also the first line of each
658 finding:
660 bzcat uploads.csv.bz2 | perl -MYAML::Syck -F, -nale '$F[-1]=~s/\s+\z//; push @{$Seen{$F[-1]}||=[]},$_; END {for my $k (keys %Seen){ delete $Seen{$k} if @{$Seen{$k}}==1; } print YAML::Syck::Dump(\%Seen)}'
662 2008-10-31 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
664 * memory leak in the syncher? It currently weighs 100M.
666 Update 2008-11-02:
668 root 10705 1.0 4.9 80192 76596 pts/32 S+ Nov02 24:05 /home/src/perl/repoperls/installed-perls/perl/pVNtS9N/perl-5.8.0@32642/bin/perl -Ilib /home/k/sources/CPAN/GIT/trunk/bin/testing-rmirror.pl
671 2008-10-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
673 * lookup by epoch and by path and use this ability on the pause to never
674 again register a file twice that doesn't need it. Let's call it
675 contains().
677 * after the dirtymark is done: fill up recentfiles with fake (historic)
678 entries; fill up with individual corrections; algorithm maybe to be done
679 with bigfloat so that we can always place something in the middle
680 between two entries. Before we must switch to bigfloat we could try to
681 use Data::Float::nextup to get the.
683 * Inotify2 on an arbitrary tree and then play with that instead of PAUSE
684 directly.
686 * dirtymark now lives in Recentfile, needs to be used in rmirror.
688 * find out why the downloader died after a couple of hours without a net
689 connection. Write a test that survives the not-existence of the other
690 end forever.
692 2008-10-15 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
694 * reconsider the HTTP epoch only. Not the whole thing over HTTP because
695 it makes less sense with tight coupling for secondary files. But asking
696 the server what the current epoch is might be cheaper on HTTP than on
697 rsync. (Needs to be evaluated)
699 * remove the 0.00 from the verbose overview in the Merged column in the
700 Z row. DONE
702 * write tests that expose the problems of the last few days: cascading
703 client/server roles, tight coupling for secondary RFs, deletes after
704 copies.
706 * Some day we might want to have policy options for the slave:
707 tight/loose/no coupling with upstream for secondary RFs. tight is what
708 we have now. loose would wait until a gap occurs that can be closed.
710 2008-10-14 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
712 * revisit all $rfs->[$i+1] places if they now make sense still
714 2008-10-11 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
716 * another bug is the fact that the mirror command deletes files before
717 it unhides the index file, thus confusing downstream slaves. We must not
718 delete before unhiding and must delete after unhiding. FIXED.
720 * new complication about the slave that is playing a server role.
721 Currently we mirror from newest to oldest with a hidden temporary file
722 as index. And when one file is finished, we unhide the index file.
723 Imagine the cascading server/slave is dead for a day. It then starts
724 mirroring again with the freshest thing and unhides the freshest index
725 file when it has worked through it. In that moment it exposes a time
726 hole. Because it now works on the second recentfile which is still
727 hidden.
729 We currently do nothing special to converge after such a drop out. At
730 least not intentionally and robustly and thought through.
732 The algorithm we use to seed the next file needs quite a lot of more
733 robustness than it currently has. Something to do with looking at the
734 merged element of the next rf and when it has dropped off, we seed
735 immediately. And if it ramains dropped off, we seed again, of course.
737 Nope, looking from smaller to larger RFS we look at the merged element
738 of this RF and at the minmax/max element of the next RF. If that
739 $rf[next]->{minmax}{max} >= $rf[this]->{merged}{epoch}, then we can stop
740 seeding it.
742 And we need a public accessor seed and unseed or seeded. But not the mix
743 of public and private stuff that then is used behind the back.
745 And then the secondary* stuff must go.
747 And we must understand what the impact is on the DONE system. Can it go
748 unnoticed that there was a hole? And could the DONE system have decided
749 the hole is covered? This should be testable with three directories where
750 the middle stops working for a while. Done->merge is suspicious, we must
751 stop it from merging non-conflatable neighbors due to broken continuity.
753 FIXED
755 2008-10-10 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
757 * Slaven suggests to have the current epoch or the whole current
758 recentfile available from the HTTP server and take it away with
759 keepalive. This direction goes the granularity down to subseconds.
761 We might want to rewrite everything to factor out transport and allow
762 the whole thing to run via HTTP.
764 2008-10-09 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
766 * are we sure we do NOT LEAVE DOT FILES around? Especially on the
767 symlink?
769 * smoker on k81 fetching from k75 to verify cascading works. See
770 2008-07-17 in upgradexxx and rsync-over-recentfile-3.pl.
772 * maybe the loop should wait for CHECKSUMS file after every upload. And
773 CPAN.pm needs to deal with timestamps in the future.
775 * do not forget the dirtymark!
777 Text: have a new flag on recentfiles with the meaning: if this
778 changes, you're required to run a full rsync over all the files. The
779 reason why we set it would probably be: some foul happened. we injected
780 files in arbitrary places or didn't inject them although they changed.
781 The content of the flag? Timestamp? The relation between the
782 recentfiles would have to be inheritance from the principal, because any
783 out of band changes would soon later propagate to the next recentfile.
785 By upping the flag often one can easily ruin the slaves.
787 last out of band change? dirtymark?
789 Anyway, this implies that we read a potentially existing recentfile
790 before we write one.
792 And it implies that we have an eventloop that keeps us busy in 2-3
793 cycles, one for current stuff (tight loop) and one for the recentfiles
794 (cascade when principal has changed), one for the old stuff after a
795 dirtymark change.
797 And it implies that the out-of-band change in any of the recentfiles
798 must have a lock on the principal file and there is the place to set the
799 dirtymark.
801 * start a FAQ, especially quick start guide questions. Also to aid those
802 problematic areas where we have no good solution, like the "links"
803 option to rsync.
805 * wish feedback when we are slow.
807 * reduce mccabe
809 * Remove a few DEBUG statements.
811 * The multiple-rrr way of doing things needs a new option to rmirror,
812 like piecemeal or so. Not urgent because after the first pass through,
813 things run smoothely. It's only ugly during the first pass.
815 * I have the suspicion that the code is broken that decides if the
816 neighboring RF needs to be seeded. I fear when too much time has gone
817 between two calls (in our case more than one hour), it would not seed
818 the neighbor. Of course this will never be noticed, so we need a good
819 test for it.
821 * local/localroot confusion: I currently pass both options but one must
822 do.
824 * accounts for early birds on PAUSE rsync daemon.
826 * hardcoded 20 seconds
828 * who mirrors the index? DOING now.
830 * which CPAN mirrors offer rsync?
832 * visit all XXX, visit all _float places
834 * rename the pathdb stuff, it's too confusing. No idea how.
836 * rrr-inotify, backpan, rrr-register
838 2008-10-08 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
840 * current bugs: the pathdb seems to get no reset, the seeding of the
841 secondaryttl stuff seems not to have an effect. Have helped myself with
842 a rand(10), need to fix this back. So not checked in. Does the rand
843 thing even help?
845 The rand thing helps. The secondaryttl stuff was in the wrong line,
846 fixed now.
848 The pathdb stuff was because I called either _pathdb or __pathdb on the
849 wrong object. FIXED now.
851 * It's not so beautiful if we never fetch the recentfiles that are not
852 the principal, even if this is correct behaviour. We really do not need
853 them after we have fetched the whole content.
855 OK, we want a switch for that: secondaryttl DONE
857 2008-10-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
859 * bug: rrr-news --max does not count correctly. with "35" it shows me 35
860 lines but with 36 it shows 110. First it repeats 35, gives 70, and then
861 it lets 40 follow. FIXED
863 * See that the long running process really only updates the principal
864 file unless it has missed a timespan during which something happened. If
865 nothing happened, it must notice even when it misses the timespan. DONE
867 * we must throw away the pathdb when we have reached the end of Z. From
868 that moment we can have a very small pathdb because the only reason for
869 a pathdb is that we know to ignore old records in old files. We won't
870 need this pathdb again before the next full pass over the data is
871 necessary and then we will rebuild it as we go along. DONE
873 2008-10-06 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
875 * I think, Done::register_one is doing wrong in that it does not
876 conflate neighboring pieces. The covered() method cannot do this because
877 it has no recent_events array at hand. But register_one has it and could
878 do it and for some reason misses to do it (sometimes).
880 This means that the three tests I just wrote can probably not survive
881 because they test with an already broken Done structure.
883 The art now is to detect how it happens, then to reproduce, then write a
884 test, then fix it.
886 So from the logfile this is what happens: we have a good interval with
887 newest file being F1 at T1. Now remotely F1 gets a change and F2 goes on
888 top of it. Locally we now mirror F2 and open a new done interval for it.
889 Then we mirror F1 but this time with the timestamp T1b. And when we then
890 try to close the gap, we do not find T1 but instead something older. We
891 should gladly accept this older piece and this would fix this bug.
893 FIXED
895 * bug to fix: when the 1h file changes while rmirror is running, we do
896 correctly sync the new files but never switch to the 6h file but rather
897 stay in a rather quick loop that fetches the 1h file again and again.
899 Is it possible that we initialize a new object? Or does
900 get_remote_recentfile_as_tempfile overwrite something in myself?
902 Want a new option: _runstatusfile => $file which frequently dumps the
903 state of all recentfiles to a file.
905 FIXED
907 2008-10-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
909 * Todo: now teach update to verify the timestamp is about to write
910 against the previous and use _increase_a_bit if it doesn't comply with
911 strict monotony. DONE
913 * The problem of rounding. So far perl's default precision was
914 sufficient. One day it won't be. FakeFloat has an easy job when it is
915 only reading and other machines have written correctly. But when we want
916 to write a floating point number that is a bit larger than the other
917 one, then we need our own idea of precision.
919 Slaven said: just append a "1". This might be going towards the end of
920 usability too quickly. I'd like something that actually uses the decimal
921 system. Well, appending a 1 also does this but...
923 E.g. we have 1.0. nextup on this architecture is starting with
924 1.0000000000000004. So there is a gap to fill: 1,2,3. Now I have
925 taken the 1.0000000000000003 and the next user comes and the time tells
926 him 1.0 again. He has to beat my number without stepping over the
927 nextup. This is much less space than I had when I chose 1,2,3.
929 What is also irritating is that nextup is architecture dependent. The
930 128 bit guy must choose very long numbers to fit in between whereas the
931 other one with 16 bit uses larger steps. But then the algorithm is the
932 same for both, so that would be a nice thing.
934 I see two situation where we need this. One is when Time::HiRes returns
935 us a value that is <= the last entry in our recentfile. In this case
936 (let's call it the end-case) we must fill the region between that number
937 and the next higher native floating point number. The other is when we
938 inject an old file into an old recentfile (we would then also set a new
939 dirtymark). We find the integer value already taken and need a slightly
940 different one (let's call it the middle-case). The difference between
941 the two situations is that the next user will want to find something
942 higher than my number in the end-case and something lower than my number
943 in the middle case.
945 So I suggest we give the function both a value and an upper bound and it
946 calculates us a primitive middle. The upper bound in the middle-case is
947 the next integer. The upper bound on the end-case is the nextup floating
948 point number. But the latter poses another problem: if we have occupied
949 the middle m between x and nextup(x), then the nextup(m) will probably
950 not be the same as nextup(x) because some rounding will take place
951 before the nextup is calculated and when the rounding reaches the
952 nextup(x), we will end up at nextup(nextup(x)).
954 So we really need to consider the nextup and the nextdown from there and
955 then the middle and that's the number we may approach asymptotically.
956 Ugly. But DONE.
958 2008-10-03 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
960 * consider deprecating the use of RECENT.recent as a symlink. It turns
961 out to need extra hoops with the rsync options and just isn't worth it.
962 Or maybe these extra hoops are needed anyway for the rest of the tree?
963 Nope, can't be the case because not all filesystems support symlinks.
965 But before doing the large step, I'll deprecate the call of
966 get_remote_recentfile_as_tempfile with an argument. Rememberr this was
967 only introduced to resolve RECENT.recent and complicates the routine far
968 beyond what it deserves.
970 DONE. Won't deprecate RECENT.recent, just moved its handling to the
971 supervisor.
973 2008-10-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
975 * I think it's a bug that the rsync_option links must be set to true in
976 order to support RECENT.recent and that nobody cares to set it
977 automatically. Similar for ignore_link_stat_errors.
979 2008-09-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
981 * Todo: find all todos together and make a plan what is missing for a
982 release.
984 - verifytree or something like that. fsck maybe.
986 - rersyncrecent, the script itself? What it do?
988 - a way to only mirror the recentfiles without mirroring the whole
989 remote system such that people can decide to mirror only partially see
990 also 2008-08-30. .shadow-xxx directory? this also needed for a
991 filesystem that is still incomplete and might need the mirrorfiles for
992 lookup(?)
994 - long living objects that mirror again and again. Inject something
995 into ta, see how it goes over to tb.
997 - how do we continue filling up the DONE system when we use an object
998 for the second time? "fully covered" and "uptodate" or new terminology.
1000 - overview called on the wrong file should be understandable
1002 - the meta data field that must change when we fake something up so that
1003 the downstream people know they have to re-fetch everything.
1005 - how tolerant are we against missing files upstream? how do we keep
1006 track? there are legitimate cases where we did read upstream index right
1007 before a file got deleted there and then find that file as new and want
1008 it. There are other cases that are not self healing and must be tracked
1009 and bugreported.
1011 - how, exactly, do we have to deal with deletes? With rsync errors?
1013 rsync: link_stat "/id/K/KA/KARMAN/Rose-HTMLx-Form-Related-0.07.meta" (in
1014 authors) failed: No such file or directory (2)
1016 The file above is a delete in 1h and a new in file 1M and the
1017 delete in the locally running rmirror did not get propagated to the 1M
1018 object. Bug. And the consequence is a standstill.
1020 It seems that a slave that works with a file below the principal needs
1021 to merge things all the way up to get rid of later deletes. Or keep
1022 track of all deletes and skip them later. So we need a trackdeletes.pm
1023 similar to the done.pm?
1025 see also 2008-08-20 about spurious deletes that really have no add
1026 counterpart and yet they are not wrong.
1028 - consider the effect when resyncing the recentfile takes longer than
1029 the time per loop. Then we never rsync any file. We need to diagnose
1030 that and force an increase of that loop time. But when we later are fast
1031 enough again because the net has recovered, then we need to switch back
1032 to original parameters. ERm, no, it's enough to keep syncing at least
1033 one file before refetching an index file.
1035 - remember to verify that no temp files are left lying around and the
1036 signal handler
1038 - status file for not long running jobs that want to track upstream with
1039 a, say, cronjob.
1041 - revisit all XXX _float areas and study Sub::Exporter DONE
1043 - persistent DB even though we just said we do not need it. Just for
1044 extended capabilities and time savings when, for example, upstream
1045 announces a reset and we get new recentfiles and could then limit
1046 ourselves to a subset of files (those that have a changed epoch) in a
1047 first pass and would only then do the loop to verify the rest. Or
1048 something.
1050 * Todo: aggregate files should know their feed and finding the principal
1051 should be done stepwise. (?)
1053 * Todo: DESTROY thing that unlocks. Today when I left the debuggerr I
1054 left locks around. DONE
1056 2008-09-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1058 * maybe extend the _overview so that it always says if and where the
1059 last file is in the next file and where the next event in the next rf
1060 would lie. No, don't like this anymore. REJECT
1062 * take the two new redundant tests out again, only the third must
1063 survive. DONE
1065 * Todo: add a sanity check if the merged structure is really pointing to
1066 a different rf and that this different rf is larger. DONE
1068 2008-09-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1070 * now test, if they are overlapping. And test if there is a file in the
1071 next rf that would fit into this rf's interval.
1073 1h 1222324012.8474 1222322541.7963 0.4086
1074 6h 1222320411.2760 1222304207.6931 4.5010 missing overlap/gap!
1075 1d 1222320411.2760 1222238750.5071 22.6835 large overlap
1076 1W 1222313218.3626 1221708477.5829 167.9835
1078 I suspect that somebody writes a merged timestamp without having merged
1079 and then somebody else relies on it.
1081 If aggregate is running, the intervals must not be extravagated, if it
1082 is not running, there must not be bounds, the total number of events in
1083 the system must be counted and must be controlled throughout the tests.
1084 That the test required the additional update was probably nonsense,
1085 because aggregate can cut pieces too. FIXED & DONE
1087 2008-09-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1089 * rrr-aggregate seems to rewrite the RECENT file even if nothing has
1090 changed. FIXED
1092 2008-09-21 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1094 * Most apparent bug at the moment is that the recentfiles are fetched
1095 too often. Only the principal should be fetched and if it has not
1096 changed, the others should not be refetched. ATM I must admit that I'm
1097 happy that we refetch more often than needed because I can more easily
1098 fix bugs while the thing is running.
1100 * Let's say, 1220474966.19501 is a timestamp of a file that is already
1101 done but the done system does not know about it. The reason for the
1102 failure is not known and we never reach the status uptodate because of
1103 this. We must get over it.
1105 Later it turns out that the origin server had a bug somewhere.
1106 1220474966.19042 came after 1220474966.19501. Or better: it was in the
1107 array of the recentfile one position above. The bug was my own.
1109 2008-09-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1111 * There is the race condition where the server does a delete and the
1112 slave does not yet know and then tries to download it because he sees
1113 the new. So for this time window we must be more tolerant against
1114 failure. If we cannot download a file, we should just skip it and should
1115 not retry immediately. The whole system should discover the lost thing
1116 later. Keeping track with the DONE system should really be a no brainer.
1118 But there is something more: the whole filesystem is a database and the
1119 recentfiles are one possible representation of it. It's a pretty useful
1120 representation I think that's why I have implemented something around
1121 it. But for strictly local operation it has little value. For local
1122 operation we would much rather have a database. So we would enter every
1123 recentfile reading and every rsync operation and for every file the last
1124 state change and what it leads to. Then we would always ignore older
1125 records without the efforts involved with recentfiles.
1127 The database would have: path,recentepoch,rsyncedon,deletedon
1129 Oh well, not yet clear where this leads to.
1131 2008-09-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1133 * Bug: the bigloop ran into a funny endless loop after EWILHELM uploaded
1134 Module-Build. It *only* rsynced the "1h" recentfile from that moment on.
1136 * statusfile, maybe only on demand, alone to have a sharp debugging
1137 tool. It is locked and all recentfiles dump themselves into it and we
1138 can build a viewer that lets us know where we stand and what's inside.
1140 * remember: only the principal recentfile needs expiration, all others
1141 shall be expired by principal if it discovers that something has move
1142 upstream.
1144 2008-09-18 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1146 * Always check if we stringify to a higher value than in the entry
1147 before. DONE
1149 * And in covered make an additional check if we would be able to see a
1150 numerical difference between the two numbers and if we can't then switch
1151 to a different, more expensive algorithm. Do not want to be caught by
1152 floating surprises. DONE
1154 2008-09-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1156 * caching has several aspects here: we can cache the interval of the
1157 recentfile which only will change when the mtime of the file changes. We
1158 must re-mirror the recentfile when its ttl has expired. Does have_read
1159 tell you anything? It counts nothing at all. Only the mtime is
1160 interesting. The ntuple mtime, low-epoch, high-epoch. And as a separate
1161 thing the have_mirrored because it is unrelated to the mtime.
1163 * Robustness of floating point calculations! I always thought that the
1164 string calculated by the origin server for the floating representation
1165 of the epoch time is just a string. When we convert it to a number and
1166 later back to a string, the other computer might come to a different
1167 conclusion. This must not happen, we want to preserve it under any
1168 circumstances. I will have to write tests with overlong sequences that
1169 get lost in arithmetic and must see if all still works well. DONE
1171 But one fragile point remains: if one host considers a>b and the other
1172 one considers them == but no eq. To prevent this, we must probably do
1173 some extra homework. DONE
1175 2008-09-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1177 * the concept of tracking DONE needs an object per recentfile that has
1178 something like these methods:
1180 do_we_have(xxx), we_have(xxx), do_we_have_all(xxx,yyy), reset()
1182 covered() register() covered()
1184 The unclear thing is how we translate points in time into intervals. We
1185 could pass a reference to the current recent_events array when running
1186 we_have(xxx) and let the DONE object iterate over it such that it only
1187 has to store a list of intervals that can melt into each other. Ah, even
1188 passing the list together with a list of indexes seems feasiable.
1190 Or maybe ask for the inverted list?
1192 Whenever the complete array is covered by the interval we say we are
1193 fully covered and if the recentfile is not expired, we are uptodate.
1195 2008-09-07 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1197 2008-09-05 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1199 * need a way to "return" the next entry after the end of a list. When
1200 the caller says "before" or "after" we would like to know if he could
1201 cover that interval/threshold or not because this influences the effect
1202 of a newer timestamp of that recentfile. DONE with $opt{info}.
1204 2008-09-04 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1206 * one of the next things to tackle: the equivalent of csync2 -TIXU.
1208 loop implies tixu (?). Nope, something like --statefile decides. Per
1209 default we do ...?
1211 T test, I init, X including removals, U nodirtymark
1213 So we have no concept of dirtymarks, we only trust that since we are
1214 running we have observed everything steadily. But people will not let
1215 this program run forever so we must consider both startup penalty and
1216 book keeping for later runs. We keep this for later. For now we write a
1217 long running mirror that merges several intervals.
1219 2008-09-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1221 * need to speed up the 02 test, it's not clever to sleep so much. Reduce
1222 the intervals!
1224 * rersyncrecent, the script: default to one week. The name of the switch
1225 is --after. Other switches? --loop!
1227 2008-08-30 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1229 * need a switch --skip-deletes (?)
1231 * need a switch --enduser that tells us that the whole tempfile
1232 discipline is not needed when there is no downstream user. (?)
1234 Without this switch we cannot have a reasonable recent.pl that just
1235 displays the recent additions. Either we accept to download everything.
1236 Or we download temporary files without the typical rsync protocol
1237 advantages.
1239 Or maybe the switch is --tmpdir? If --tmpdir would mean: do not use
1240 File::Temp::tempdir, this might be a win.
1242 2008-08-29 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1244 * apropos missing: we have no push, we never know the downstream
1245 servers. People who know their downstream hosts and want to ascertain
1246 something will want additional methods we have never thought about, like
1247 update or delete a certain file.
1249 2008-08-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1251 * tempted to refactor rmirror into resolve_symlink, localize, etc.
1252 Curious if rsync_options=links equal 0 vs. 1 will make the expected
1253 difference.
1255 * rsync options: it's a bit of a pain that we usually need several rsync
1256 options, like compress, links, times, checksum and that there is no
1257 reasonable default except the original rsync default. I think wee can
1258 safely assume that the rsync options are shared between all recentfile
1259 instances within one recent tree.
1261 2008-08-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1263 * deletes: if a delete follows an add quickly enough it may happen that
1264 a downstream mirror did not see the add at all! It seems this needs to
1265 be mentioned somewhere. The point here is that even if the downstream is
1266 never missing the principal timeframe it may encounter a "delete" that
1267 has no complimentary "add" anywhere.
1269 2008-08-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1271 * I suspect the treat of metadata is incorrect during read or something.
1272 The bug that I am watching is that between 06:08 and 06:09 the 6h file
1273 contained more than 6 hours worth of data. At 06:08 we merged into the
1274 1d file. We need to take snapshots of the 6h file over the course of an
1275 hour or maybe only between XX:08 and XX:09? Nope, the latter is not
1276 enough.
1278 Much worse: watching the 1h file: right at the moment (at 06:35) it
1279 covers 1218867584-1219120397 which is 70 hours.
1281 Something terribly broken. BTW, 1218867584 corresponds to Sat Aug 16
1282 08:19:44 2008, that is when I checked out last time, so it seems to be
1283 aggregating and never truncating?
1285 No, correct is: it is never truncating; but wrong is: it is aggregating.
1286 It does receive a lot of events from time to time from a larger file.
1287 Somehow a large file gets merged into the small one and because the
1288 "meta/merged" attribute is missing, nobody is paying attention. I
1289 believe that I can fix this by making sure that metadata are honoured
1290 during read. DONE and test adjusted.
1292 2008-08-17 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1294 * grand renaming plan
1296 remotebase => remoteroot to fit well with localroot DONE
1297 local_path() => localroot seems to me should already work DONE
1298 recentfile_basename => rfilename no need to stress it has no slash DONE
1300 filenameroot??? Doesn't seem too bad to me today. Maybe something like
1301 kern? It would anyway need a deprecation cycle because it is an
1302 important constructor.
1304 * I like the portability that Data::Serializer brings us but the price
1305 is that some day we might find out that it is slowing us a bit. We'll
1306 see.
1308 2008-08-16 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1310 * should we not enter the interval of the principal (or the interval of
1311 the merging file?) in every aggregated/merged file?
1313 * we should aim at a first release and give up on thinking about
1314 sanitizing stuff and zloop. Let's just admit that a full traditional
1315 rsync is the only available sanitizer ATM. Otherwise it's complicated
1316 stuff: sanitizing on the origin server, sanitizing on the slaves,
1317 sanitizing forgotten files, broken timestamps, etc. Let's delay it and
1318 get the basics out before this becomes a major cause for mess.
1320 2008-08-13 Andreas Koenig <k@andreas-koenigs-computer.local>
1322 * On OSes not supporting symlinks we expect that RECENT.recent contains
1323 the contents of the principal recentfile. Actually this is identical on
1324 systems supporting symlinks. Simple, what follows from that is that we
1325 need to keep the serializer in the metadata because we cannot read it
1326 from the filename, doesn't it? Of course not. It's a chicken and egg
1327 problem. This leaves us with the problem to actually parse the
1328 serialized data to find out in which format it is. So who can do the 4
1329 or 5 magics we wanted to support? File::LibMagic?
1331 2008-08-09 Andreas Koenig <k@andreas-koenigs-computer.local>
1333 * remotebase and recentfile_basename are ugly names. Now that we need a
1334 word for the shortest/principal/driving recentfile too we should do
1335 something about it.
1337 localroot is good. rfile is good. local_path() is bad, local_path($path)
1338 is medium, filenameroot() is bad, remotebase is bad, recentfile is
1339 already deprecated.
1341 Up to now remotebase was the string that described the remote root
1342 directory in rsync notation, like pause.perl.org::authors. And
1343 recentfile_basename was "RECENT-1h.yaml".
1345 2008-08-08 Andreas Koenig <k@andreas-koenigs-computer.local>
1347 * The test that was added in today's checkin is a good start for a test
1348 of rmirror. We should have more methods in Recent.pm: verify,
1349 addmissingfiles. We should verify the current tree, then rmirror it and
1350 then verifytree the copy. We could then add some arbitrary file and let
1351 it be discovered by addmissingfiles, then rmirror again and then
1352 verifytree the copy again.
1354 Then we could start stealing from csync2 sqlite database [no port to
1355 OSX!] and fill a local DB. And methods to compare the database with the
1356 recentfiles. Our strength is that in principle we could maintain state
1357 with a single float. We have synced up to 1234567890.123456. If the Z
1358 file does not add new files all we have to do is mirror the new ones and
1359 delete the goners.
1361 This makes it clear that we should extend current protocol and declare
1362 that we cheat when we add files too late, just to help the other end
1363 keeping track. Ah yes, that's what was meant when zloop was mentioned
1364 earlier.
1366 Maybe need to revisit File::Mirror to help me with this task.
1368 2008-08-07 Andreas Koenig <k@andreas-koenigs-computer.local>
1370 * There must be an allow-me-to-truncate flag in every recentfile.
1371 Without it one could construct a sequence of updates winning the locking
1372 battle against the aggregator. Only if an aggregator has managed to
1373 merge data over to the next level, truncating can be allowed. DONE with
1374 accessor merged.
1376 2008-08-06 Andreas Koenig <k@andreas-koenigs-computer.local>
1378 * We should probably guarantee that no duplicates enter the aggregator
1379 array.
1381 2008-08-02 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1383 * To get merge operation faster would need a good benchmark test. What
1384 02 spits out isn't reliable enough and is dominated by many other
1385 things. Between
1387 commit 10176bf6b79865d4fe9f46e3857a3b8669fa7961
1388 Author: Andreas J. Koenig <k@k75.(none)>
1389 Date: Sat Aug 2 07:58:04 2008 +0200
1391 and
1393 commit 3243120a0c120aaddcd9b1f4db6689ff12ed2523
1394 Author: Andreas J. Koenig <k@k75.(none)>
1395 Date: Sat Aug 2 11:40:29 2008 +0200
1397 there was a lot of trying but the effect is hardly measurable with
1398 current tests.
1400 * overhead of connecting seems high. When setting
1401 max_files_per_connection to 1 we see that.
1403 2008-08-01 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1405 * 1217622571.0889 - 1217597432.86734 = 25138.2215600014
1407 25138.2215600014/3600 = 6.98283932222261
1409 It jumps into the eye that this is ~ 7 hours, not ~6, so there seems to
1410 be a bug in the aggregator. FIXED
1412 2008-07-27 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1414 * e.g. id/Y/YE/YEWENBIN/Emacs-PDE-0.2.16.tar.gz: Do we have it, should
1415 we have it, can we mirror it, mirror it!
1417 I fear this needs a new class which might be called
1418 File::Rsync::Mirror::Recent. It would collect all recentfiles of a kind
1419 and treat them as an entity. I realize that a single recentfile may be
1420 sufficient for certain tasks and that it is handy for the low level
1421 programmer but it is not nice to use. If there is a delete in the 1h
1422 file then the 6h file still contains it. Seekers of the best information
1423 need to combine at least some of the recentfiles most of the time.
1425 There is the place for the Z loop!
1427 But the combination is something to collect in a database, isn't it. Did
1428 csync2 just harrumph?
1430 2008-07-26 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1432 * it just occurred to me that hosts in the same mirroring pool could
1433 help out each other even without rewriting the recentfile. Just fetch
1434 the stuff to mirror from several places, bingo. But that's something
1435 that should rather live in a separate package or in rsync directly.
1437 * cronjobs are unsuited because with ntp they would all come at the full
1438 minute and disturb each other. Besides that I'd hate to have a backbone
1439 with more than a few seconds latency.
1441 2008-07-25 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1443 * a second rsync server with access control for PAUSE. Port? 873 is the
1444 standard port, let's take 8873.
1446 * if there were a filesystem based on this, it would have a slow access
1447 to inexistent files. It would probably provide wrong readdir (only based
1448 on current content) or also a slow one (based on a recentfile written
1449 after the call). But it would provide fast access to existing files. Or
1450 one would deliberately allow slightly blurred answers based on some
1451 sqlite reflection of the recentfiles.
1453 * todo: write a variant of mirror() that combines two or more
1454 recentfiles and treats them like one
1456 * todo: signal handler to remove the tempfile
1458 2008-07-24 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1460 * now that we have the symlink I forgot how it should be used in
1461 practice.
1463 * the z loop: add missing files to Z file. Just append them (instead of
1464 prepending). So one guy prepends something from the Y file from time to
1465 time and another guy appends something rather frequently. Collecting
1466 pond. When Y merges into Z, things get epoch and the collecting pond
1467 gets smaller. What exactly are "missing files"?
1469 take note of current epoch of the alpha file, let's call it the
1470 recent-ts
1472 find all files on disk
1474 remove all files registered in the recentworld up to recent-ts
1476 remove all files that have been deleted after recent-ts according to
1477 recentworld
1479 2008-07-23 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1481 * rersyncrecent might be a cronjob with a (locked) state file which
1482 contains things like after and maybe last z sync or such?
1484 rrr-mirror might be an alternative name but how would we justify the
1485 three Rs when there is no Re-Rsync-Recent?
1487 With the --loop parameter it is an endless loop, without it is no loop.
1488 At least this is simple.
1490 * todo: new accssor z-interval specifies how often the Z file is updated
1491 against the filesystem. We probably want no epoch stamp on these
1492 entries. And we want to be able to filter the entries (e.g. no
1493 by-modules and by-category tree)
1495 2008-07-20 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1497 * Fill the Z file. gc or fsck or both. Somehow we must get the old files
1498 into Z. We do not need the other files filled up with filesystem
1499 contents though.
1501 * need interface to query for a file in order to NOT call update on
1502 PAUSE a second time within a short time.
1504 2008-07-19 Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de>
1506 * recommended update interval? Makes no sense, is different for
1507 different users.
1509 * Moosify
1511 Local Variables:
1512 mode: change-log
1513 change-log-default-name: "Todo"
1514 tab-width: 2
1515 left-margin: 2
1516 End: